Dataset
A Dataset is made up of Data Objects which can have features with various types.
Types of Datasets
Datasets with one attribute are univariate. Datasets with more than one attribute are called Multivariate Dataset, commonly given in a table
There is also bivariate and trivariate data but this description isn't being used a lot.
- Data Streams, sensor data
- Time Series Data
- Social Network, Graph
- Hierarchy
- Relationenmodell
- Heterogeneous databases and legacy databases
- NoSQL
- Spatial Data (also spatial-temporal)
- Multimedia (video, etc.)
- Text (term frequency vectors) → vocab with occureances for each word for every document → Natural Language Processing
- WWW
- Data Warehouse System - Architektur
- Transactional Database e.g. customer bought items for each customer (Pattern Mining, which articles are bought together)
- Data Matrix (without column names)
Ordered Data:
- video
- Time Series Data
- sequential
- genetic sequence data
Spatial, image, multimedia:
- maps
- image
- video
Also see Curse of Dimensionality.
Proximity Measure for differenty attribute types:
- Nominal → Simple Matching Coefficient, One-Hot-Encoding
- Ordinal → Integer Encoding
- Binary → Contingency table for binary data